The further back you look, the further forward you see. (Winston Churchill)

To my reckoning, there have been three cycles of the 'Centralized versus Decentralized Computing' debate over the last 30-odd years.

It's one reason why 'ageism' in the software industry (and others) can be translated into 'short-sightedness'.

I noticed the centralization swings back when I was a physics student, and came up with the idea that it basically goes back and forth with a period of about seven years. In that seven years, they forget why they switched away and all that pain is a distant memory. Different people are part of the group, and maybe some of the new kids have these hot ideas they learned in school, read about on Digg, or whatever. It's time to upgrade everything and maybe try something new. Everything new, however, is old.

People have to give up something no matter which way they go. If they go for central control, then everyone gets the same thing and some people lose the power to configure things just the way they like them. If they control it themselves, then they have to spend time managing it themselves.

The problem comes when they don't realize why they decided one way or the other. After several years of centralization, they start griping that they need someone to manage all this stuff for them (or need software to do it, and so on). They figure that they'll save a lot of time by unloading the wasteful computer work everyone does for themselves. They do that and they are happy for a while. Then, several years later, they start thinking that they could be a lot more efficient if they could customize their computers since that centralization policy tries to make everyone do the same thing. They de-centralize control, and everyone is happy for a few years, until they realize they are doing a lot of their own computer support. They decide to centralize again...

I was struck by the same thought recently. In a class we were briefly introduced to Grid computing. I don't know much about Grid computing, this was just a very brief intro. But the picture I formed in my mind was amusing. It's supposed to be the new, up and coming thing. All I could think of at the time were old scifi movie scenes of huge computer rooms with huge systems and many companies calling in at the same time.

I know it makes sense and in today's world Grid computing is not the old "huge system" we used to see in the movies, but as Grey Fox says, it does seem that we're going round in circles. We started with a central computer area that many companies shared, and now we're talking about going back to the exact same configuration. Granted that new configuration has huge improvements over the old.

Maybe it's me but I see a huge difference between hardware centralization and control centralization. I could really care less about hardware centralization. Mainframe, mid-range, mini's or clusters of cheap cpus - don't care. Just as long as it stays a bazaar and I never have to go back to dropping off punch cards or picking up reams of 11x22 fanfold paper. To me, that's where the circle always needed to be broken - the Data Center Brahman need not re-apply.

It's been my experience that as the hardware goes so goes the company. Most companies that have taken to a Centralized data processing scheme, tend to do it to enforce standards across the company thus limiting the possible solutions. What then happens is it takes so long to implement a standard solution that user communities tend to go rogue, and start setting up ther own independent programs.
-- Grey Fox
"We are gray. We stand between the darkness and the light" B5